Journi
Agentic AI itinerary engine generating personalized trips in under 30s
Overview
What this is
Journi is an AI trip planner built around an agentic itinerary engine that autonomously generates personalized, editable day-to-day travel plans.
Problem
Why it needed building
A genuinely useful itinerary means pulling live data from several unrelated services — flights and hotels, weather, maps — and reasoning over all of it together, not just calling one API and templating the response.
Solution
How it works
Built an agentic engine on top of Grok's API using tool-use and function calling, orchestrating real-time calls across 3+ third-party services to assemble a full itinerary in under 30 seconds. A Redis-backed job queue handles the 10-30 second generation time asynchronously, so it never blocks the request cycle.
Architecture
How it's put together
- →Agentic itinerary engine built on Grok's API using tool-use / function calling to orchestrate flights/hotels, weather, and maps data
- →Redis-backed job queue decoupling 10-30 second AI generation tasks from the request-response cycle
- →Stripe subscription billing with a 2-tier (free/paid) pricing model
- →Usage limits enforced at the API layer via Clerk-authenticated sessions
Challenges
The hard part
Coordinating multiple third-party API calls inside a single agentic reasoning loop, without the whole generation stalling if one service — say, weather — is slow to respond.
Trade-offs
What I gave up on purpose
Offloading generation to a job queue means the frontend has to poll or subscribe for a result instead of getting an instant response — a deliberate trade of raw request latency for not blocking the server on a 30-second AI call.
Performance
What changed, measurably
Generates personalized, editable itineraries in under 30 seconds, with 5+ full trips generated to date.
Next
What I'd build next
Add real-time collaborative editing so a group can adjust a shared itinerary together instead of each person holding their own editable copy.